An interactive PDF reader, that utilizes Semantic Scholar's data and corpus to provide a novel reading experience to researchers. More to come soon 🎉.
- Install NodeJS.
To launch the development version of this server, first install the dependencies by running:
npm installThen you can launch the app as follows:
npm startA web browser should open automatically to open the app. You won't see much interesting at first, as no PDF has been loaded. Load a PDF by visiting this URL.
You can in theory provide the URL for any PDF as the value
of the file query parameter. That said, at the time this
README was written, there was only a few PDFs for which the
augmented reading experience is available.
If you'd like to view a list of all available papers, visit this URL.
We leverage Mozilla's PDF.js for rendering the PDF in a browser. We maintain a fork with several changes specific to our application. The fork is linked to the code as a submodule.
The code in public/ is produced from our fork, and serves
as a base into which our additional interface elements are
rendered. You can regenerate this code and overwrite the copies
in public/ by executing these commands:
git submodule init
git submodule update
bash bin/build_pdf.js.sh
Of note, make sure you restore the HEAP analytics tracking snippet
after regenerating viewer.html. It should be easy to identify in the
resulting git diff.
The code for the enhancements we add to the PDF is present in
src/. We've elected to use TypeScript
and React for writing the interface.
By default the UI proxies API requests to the production API (https://s2-reader.apps.allenai.org). This makes it possible to run a self-standing version of the UI locally.
To use a local version of the API, export the PROXY environment
variable like so:
export PROXY='http://localhost:3000'See the README in the api/ sibling directory for more information
about running it locally.